The integration of digital technologies in SCADA systems has made it easier to monitor and automate the industrial process, but at the same time, it has made the OT environment vulnerable to cyberattacks. The OT environment based on SCADA systems has unique operational requirements, has heterogeneous devices, continuous flow of data, and also has safety considerations. As a result, it becomes hard to detect any threats to the environment in a timely manner. Conventional approaches of intrusion detection rely heavily on predefined rules, signatures, and static behavioral models that might not be enough to identify unknown attacks and behavioral changes in devices and networks. In this research paper, an approach for detecting cyber threats using behavioral analytics through machine learning in SCADA-based OT environments has been suggested using a combination of Liquid Neural Networks (LNNs), Hypergraph Neural Networks (HGNNs), and Deep Evidential Learning (EDL). The LNN is employed to capture the temporal changes in the behavior of the system through its operations and data flows. The HGNN component is used to capture complex dependencies between devices, communication activities, network activities, and other interconnected entities in the OT environment, where conventional pairwise representation may fail to model dependencies between several entities. Proposed framework examines behavioral patterns in SCADA-related datasets and uses temporal patterns, high order dependencies, and evidential information to separate normal operational behavior from any potentially malicious behavior. Evaluation of the proposed framework provided 97.29% accuracy in identifying cyber threat patterns in the investigated SCADA-based OT environment